Re: Gentoo with ext4-patch-queue snapshots

[Mingming Cao <cmm@us.ibm.com>]

[-- Attachment #1: Type: text/plain, Size: 3279 bytes --]


On Wed, 2008-07-02 at 10:19 -0700, Mingming Cao wrote:
> On Tue, 2008-07-01 at 17:50 +0000, Gary Hawco wrote:
> > Mingming,
> > 
> > Can you post that patch somewhere for download? I access my email using
> > Windows Vista, not in linux, so it would be very laborious to hand copy
> > this patch and recreate it in linux.
> > 
> Patch attached. 


Please use this patch instead, after discuss with Ted, I found an issue
with the patch I sent to the list. ext4 patch queue is also updated with
latest patch.

Ext4: fix delalloc i_disksize early update issue

From: Mingming Cao <cmm@us.ibm.com>

Ext4_da_write_end() uses ext4_bh_unmapped_or_delay() function to check
if it extend the file size without need for allocation. But at that time
the buffer has not being dirtied yet (done in code later in
block_commit_write()), so it always return true and update i_disksize
(before block allocation). we could fix that ext4_da_write_end() to not
use this helper function.

This also fixed another issue:
The i_disksize is updated at ext4_da_write_end() time if
writes to the end of file, and the buffers are all have
blocks allocated. But in the case blocksize < pagesize, and
if the page has, say, the first buffer marked
as buffer_delay, and the write is to EOF and on the third buffer, which
has block already allocated, we certainly need to extend the i_disksize.


Signed-off-by: Mingming Cao <cmm@us.ibm.com>
---
 fs/ext4/inode.c |   31 +++++++++++++++++++++++++++++--
 1 file changed, 29 insertions(+), 2 deletions(-)

Index: linux-2.6.26-rc8/fs/ext4/inode.c
===================================================================
--- linux-2.6.26-rc8.orig/fs/ext4/inode.c	2008-07-02 09:53:42.000000000 -0700
+++ linux-2.6.26-rc8/fs/ext4/inode.c	2008-07-02 13:22:52.000000000 -0700
@@ -1891,6 +1891,32 @@ out:
 	return ret;
 }
 
+/*
+ * Check if we should update i_disksize
+ * when write to the end of file but not require block allocation
+ */
+static int ext4_da_should_update_i_disksize(struct page *page,
+					 unsigned long offset)
+{
+	struct buffer_head *head, *bh;
+	unsigned int curr_off = 0;
+
+	head = page_buffers(page);
+	bh = head;
+	do {
+		unsigned int next_off = curr_off + bh->b_size;
+
+		if (curr_off <= offset && offset < next_off)
+		   if (!buffer_mapped(bh) || (buffer_delay(bh)))
+			return 0;
+		   else
+			return 1;
+		curr_off = next_off;
+	} while ((bh = bh->b_this_page) != head);
+
+	return 1;
+}
+
 static int ext4_da_write_end(struct file *file,
 				struct address_space *mapping,
 				loff_t pos, unsigned len, unsigned copied,
@@ -1900,6 +1926,10 @@ static int ext4_da_write_end(struct file
 	int ret = 0, ret2;
 	handle_t *handle = ext4_journal_current_handle();
 	loff_t new_i_size;
+	unsigned long start, end;
+
+	start = pos & (PAGE_CACHE_SIZE - 1);
+	end = start + copied;
 
 	/*
 	 * generic_write_end() will run mark_inode_dirty() if i_size
@@ -1909,8 +1939,7 @@ static int ext4_da_write_end(struct file
 
 	new_i_size = pos + copied;
 	if (new_i_size > EXT4_I(inode)->i_disksize)
-		if (!walk_page_buffers(NULL, page_buffers(page),
-				       0, len, NULL, ext4_bh_unmapped_or_delay)){
+		if (ext4_da_should_update_i_disksize(page, end)) {
 			/*
 			 * Updating i_disksize when extending file without
 			 * need block allocation


[-- Attachment #2: delalloc_i_disksize_update-fix.patch --]
[-- Type: text/x-patch, Size: 2703 bytes --]

Ext4: fix delalloc i_disksize early update issue

From: Mingming Cao <cmm@us.ibm.com>

Ext4_da_write_end() uses ext4_bh_unmapped_or_delay() function to check
if it extend the file size without need for allocation. But at that time
the buffer has not being dirtied yet (done in code later in
block_commit_write()), so it always return true and update i_disksize
(before block allocation). we could fix that ext4_da_write_end() to not
use this helper function.

This also fixed another issue:
The i_disksize is updated at ext4_da_write_end() time if
writes to the end of file, and the buffers are all have
blocks allocated. But if the page had one buffer marked
as buffer_delay, and the write is at EOF and on a buffer has block
already allocated, we certainly need to extend the i_disksize.

Signed-off-by: Mingming Cao <cmm@us.ibm.com>
---
 fs/ext4/inode.c |   31 +++++++++++++++++++++++++++++--
 1 file changed, 29 insertions(+), 2 deletions(-)

Index: linux-2.6.26-rc8/fs/ext4/inode.c
===================================================================
--- linux-2.6.26-rc8.orig/fs/ext4/inode.c	2008-07-02 09:53:42.000000000 -0700
+++ linux-2.6.26-rc8/fs/ext4/inode.c	2008-07-02 13:22:52.000000000 -0700
@@ -1891,6 +1891,32 @@ out:
 	return ret;
 }
 
+/*
+ * Check if we should update i_disksize
+ * when write to the end of file but not require block allocation
+ */
+static int ext4_da_should_update_i_disksize(struct page *page,
+					 unsigned long offset)
+{
+	struct buffer_head *head, *bh;
+	unsigned int curr_off = 0;
+
+	head = page_buffers(page);
+	bh = head;
+	do {
+		unsigned int next_off = curr_off + bh->b_size;
+
+		if (curr_off <= offset && offset < next_off)
+		   if (!buffer_mapped(bh) || (buffer_delay(bh)))
+			return 0;
+		   else
+			return 1;
+		curr_off = next_off;
+	} while ((bh = bh->b_this_page) != head);
+
+	return 1;
+}
+
 static int ext4_da_write_end(struct file *file,
 				struct address_space *mapping,
 				loff_t pos, unsigned len, unsigned copied,
@@ -1900,6 +1926,10 @@ static int ext4_da_write_end(struct file
 	int ret = 0, ret2;
 	handle_t *handle = ext4_journal_current_handle();
 	loff_t new_i_size;
+	unsigned long start, end;
+
+	start = pos & (PAGE_CACHE_SIZE - 1);
+	end = start + copied;
 
 	/*
 	 * generic_write_end() will run mark_inode_dirty() if i_size
@@ -1909,8 +1939,7 @@ static int ext4_da_write_end(struct file
 
 	new_i_size = pos + copied;
 	if (new_i_size > EXT4_I(inode)->i_disksize)
-		if (!walk_page_buffers(NULL, page_buffers(page),
-				       0, len, NULL, ext4_bh_unmapped_or_delay)){
+		if (ext4_da_should_update_i_disksize(page, end)) {
 			/*
 			 * Updating i_disksize when extending file without
 			 * need block allocation